Generalized Extended Stochastic Gradient Algorithm Implemented Parameter Identification for Complex Multivariable-Systems

摘要

In this paper, a new recognition method is deduced based on the theory of model equivalence in order to modify the parameter estimation for the multi-input nonlinear equation-error autoregressive moving average(Multi-variable) system. Using the theory of model equivalence, using the auxiliary model to handle the colored noise, the proposed algorithm reduces the number of unknown noise items in the recognition model information vector and achieves better recognition accuracy. For comparison, we use the recursive generalized extended least squares (RGELS) algorithm. To confirm the effectiveness of the algorithm, an example is shown.